Virtual Linux Server Disaster Recovery Planning...infrastructure to restart operations after a disaster. While many of today's larger computer systems contain built-in programs for

Virtual Linux ServerDisaster Recovery PlanningRick BarlowNationwide Insurance

August 10, 2011

This information is for sharing only and not an endorsement by Nationwide Insurance2

Agenda

• Definitions• Our Environment• Business Recovery Philosophy at Nationwide• Planning• Execution


Definitions

• High Availability

– “With any IT system it is desirable that the system and its components (be they hardware or software) are up and running and fully functional for as long as possible, at their highest availability. The most desirable high availability rate is known as “five 9s”™, or 99.999% availability. A good deal of planning for high availability centers around backup and failover processing and data storage and access.”

– Deal with significant outage within data center• LPAR failure

• Operating System outage

• Application ABEND


High AvailabilityLPAR

1LPAR

2LPAR

6

Cisco Switchw/ Firewall

Internet


LPAR3

LPAR4

LPAR8

OSA OSA OSA OSA OSA OSA OSA OSA OSA OSA OSA OSA

intranet





OSA OSA OSA OSA

Front End Back End Tools

vSwitchvSwitch vSwitch vSwitch

vSwitch vSwitch vSwitchvSwitch vSwitch vSwitch

vSwitch vSwitch

VLAN 1

VLAN 2

HTTPServer

HTTPServer

WASNode Session

Node

WASNode WAS

Node

VLAN 3WASNode WAS

Node

WASNode Session

Node

VLAN 4DBServer

DBServer

VLAN 5DBServer

DBServer

VMTCPIP

VMTCPIP

VMTCPIP

VMTCPIP

VMTCPIP

VMTCPIP

Production


Definitions

• Disaster Recovery

– “Disaster recovery in information technology is the ability of an infrastructure to restart operations after a disaster. While many of today's larger computer systems contain built-in programs for disaster recovery, standalone recovery programs often provide enhanced features. Disaster recovery is used both in the context of data loss prevention and data recovery.”

– Deal with complete outage• Natural catastrophe

• Data center

• Major hardware failure


Our Environment

• Four z196 installed in 2011Current configuration:

– Production boxes• 36 IFLs

• 404GB memory

• 6 z/VM LPARs (plus system programmer configuration test)

• Tier 4+ data center

– Fully redundant power, telecom, generators, etc

– Development boxes• 21 IFL engines on development box

• 663GB memory

• 6 z/VM LPARs (plus system programmer test)


Planning

• Design• Challenges• Priorities• Setup• Automation• Documentation• Teamwork


Design

• What– Identify what needs to be recovered

• Everything or subset?• Priorities – recovery order

• When– Need to know recovery objectives

• Where– Identify where recovery will occur

• Second site• Vendor site

• How– Identify how to transfer programs and data– Identify how to perform recovery


Challenges

• Production configuration changes may require DR configuration changes– Processor model changes (not necessarily)– Capacity– Network configuration– Transition

• Consistent distribution and synchronization of boot and DR scripts

• Commitment to regular testing• DR Infrastructure ≠ Application DR


Priorities

• Prioritizing your application recovery must be done by the people in your organization that understand the business processes.

• Business requirements that drive recoverytime-frame– regulated– financial / investments– responsive to customers


Setup

• Asynchronous replication with SRDF between production and recovery sites

• Replicated volumes at recovery site– z/VM different unit address between prod and dev/test/dr– SAN target WWPN and LUN differ between production and

recovery• Changes for recovery are automated in Linux script run at boot

• Manual processes– Initiate Clone copies– Vary ECKD DR volumes online– Start Linux servers– Update DNS to reflect DR IP addresses for all servers– Middleware and application hard-coded parameters (e.g. IP

addresses)


Setup

• The DR process would be the same for the following failures– System z failure– DASD / SAN storage frame failure

• Long distance (inter-date center) fiber failure– No DR required, (redundant links installed)


Setup

ReplicationFibre

FICON(x8)IBM

z196

IBMz196

EMC V-MaxStorageECKD

EMC V-MaxStorageECKD

OpenSystems

SAN

OpenSystems

SAN

FCP (x12)

FCP (x12)Fibre

Prod

Dev

SRDF Replication

FICON(x8)

ReplicationFibre

IBMz196

IBMz196


Normal Production

Production

SAN SAN

DMX DMX

Fiber

Fiber

Replication

ReplicationReplication

R1 R2

DR


Failure Happens

Production

SAN SAN

DMX DMX

Fiber

Fiber

DRIf failure occurs: manually stop replication; initiate DR Clones; Bring ECKD volumes on-line at DR site, start Linux servers


Recovery Begins

Production

SAN SAN

V-Max V-Max

Fiber

Fiber

DRServers identify DR configuration, change IP address, change SAN parameters, register new IP with DNS, start replication south to north.


Recovery Begins

Production

SAN SAN

V-Max V-Max

Fiber

Fiber

DR

Replication

R2 R1

Servers identify DR configuration, change IP address, change SAN parameters, register new IP with DNS, start replication south to north.


Automation

• Avoid manual processes– Dependence on key individuals

– Prone to mistakes

– Slow

• Automated processes– Requires only basic knowledge of environment and technologies in use

– Accuracy

– Repeatable

– Faster

– Does not mean build it once then ignore;Requires regular review and updates


Automation

• Automation begins at provisioning

– DR configuration stored with production configuration

– CMS NAMES file• Contains all information about provisioned server

• Copy stored on DR disk also

• Also used to generate report of server definitions for easy lookup

– Linux PARM file stored on CMS disk• Stored on disk accessible at boot time

• Copy stored on DR disk also

– Define everything needed to provision server and at boot time


Automation

• Extract from LINUX NAMES file forone guest

:nick.WS001:userid.PZVWS001:node.VN2:desc.Prod web server 1:env.PROD:hostname.PZVMWS001:load.LINUXWS:ip.10.1.1.1:vswitch.PRODVSW1:vlan.2102:ip_nb.10.2.1.1 :vsw_nb.NETBKUP1:vlan_nb.3940

:ip_dr.10.221.1.1:vsw_dr.PRODVSW1:vlan_dr.2102:ip_drbu.10.222.1.1:vsw_drbu.NETBKUP1:vlan_drbu.3940:oth_ip.10.1.1.5 10.1.1.15:dr_oth_ip.10.221.1.5 10.221.1.15:status.2005-09-08:gold.V1.2:memory.256M:cpus.1:share.200,LS-200:comments.:storage.2.3G:storage_os.7.1G:bootdev.251


Automation

• Extract from LINUX NAMES file for one guest (cont’d):storage_san.16.86G :sanluns.R1:0100:5006048AD52D2588:0059000000000000:8.43 R1:0100:5006048AD52D2588:005A000000000000:8.43 R1:0200:5006048AD52D2587:0059000000000000:8.43 R1:0200:5006048AD52D2587:005A000000000000:8.43

:storage_san_dr.16.86G :sanluns_dr.R2:0100:5006048AD52E4F87:006A000000000000:8.43 R2:0100:5006048AD52E4F87:006B000000000000:8.43 R2:0200:5006048AD52E4F88:006A000000000000:8.43 R2:0200:5006048AD52E4F88:006B000000000000:8.43


Automation

• PARM fileHOST=pzvmws001ADMIN=10.1.1.1BCKUP=10.2.1.1DRADMIN=10.221.1.1DRBCKUP=10.222.1.1ENV=PRODDRVIP=10.1.1.5,10.1.1.15BOOTDEV=251VIP=10.221.1.5,10.221.1.15SAN_1=0100:5006048AD52D2588:005A000000000000,0200:5006048AD52D2587:005A000000000000SAN_2=0100:5006048AD52D2588:0059000000000000,0200:5006048AD52D2587:0059000000000000SAN_3=0100:5006048AD52E4F87:006A000000000000,0200:5006048AD52E4F88:006A000000000000SAN_4=0100:5006048AD52E4F87:006B000000000000,0200:5006048AD52E4F88:006B000000000000


Automation

• Alternate Start-up Scripts

– Identify production or DR mode• VMCP – interact with CP

• CMSFS – read CMS files

– Set parameters for environment• Hostname to /etc/HOSTNAME

• IP addresses to /etc/sysconfig/network/ifcfg-qeth-bus-ccw-0.0.xxxx

• SAN LUN information

• Color prompt by environment– Prod = Red

– DR = Yellow

– Tools = Green


NZVWS001 AT VN1VN1N

Automation

• Extract from boot.config

# Setup variablesecho "1" > /sys/bus/ccw/devices/0.0.0191/onlinesleep 5# modprobe required just in casemodprobe cpintPARMDEV=`grep 191 /proc/dasd/devices|awk '{print $7}'`QUSERID=`hcp query userid`GUEST=ècho $QUSERID|cut -d" " -f 1`LOCO=ècho $QUSERID|cut -c14`LPAR=ècho $QUSERID|cut -c13-15`BOX=ècho $QUSERID|cut -c1`cmsfscat -d /dev/$PARMDEV -a ${GUEST}.PARMFILE > /tmp/sourceinfo. /tmp/sourceinfo

echo "0" > /sys/bus/ccw/devices/0.0.0191/online

NZVWS001N


Automation

• Result of cmsfscat

cat /tmp/sourceinfoHOST=pzvmws001ADMIN=10.1.1.1BCKUP=10.2.1.1DRADMIN=10.221.1.1DRBCKUP=10.222.1.1ENV=PRODDRVIP=10.1.1.5,10.1.1.15BOOTDEV=251VIP=10.221.1.5,10.221.1.15SAN_1=0100:5006048AD52D2588:005A000000000000,0200:5006048AD52D2587:005A000000000000SAN_2=0100:5006048AD52D2588:0059000000000000,0200:5006048AD52D2587:0059000000000000SAN_3=0100:5006048AD52E4F87:006A000000000000,0200:5006048AD52E4F88:006A000000000000SAN_4=0100:5006048AD52E4F87:006B000000000000,0200:5006048AD52E4F88:006B000000000000


Automation

ST)CLR="44"; #Blue;;

PT)CLR="45"; #Purple;;

UAT | IT)CLR="46"; #Turq;;

*)CLR="42"; #GreenENV="UNK";;;

esac

• More extract from boot.config

case "$ENV" inPROD)if [ "$LOCO" == "$BOX" ]

thenCLR="41"; #Red

elseCLR="43"; #Yellow/GoldENV="DR";

fi;;

DEV | JT | TOOLS | TOOL)CLR="42"; #Green;;

Examples:barlowr@szvmjt002:JT:barlowr>barlowr@nzvmws001:PROD:barlowr>


Documentation

• Document everything– Declaration criteria– Contact information

• Operating System• Middleware• Application• Network• Security

– Lists of servers– Recovery process– Verification process– Fail-back process


Documentation

• DR Procedure:– Confirm DISASTER declaration

– Begin shutdown all test/development guests to insure sufficient capacity.

– Bring up production DR guests identified by business units for each application environment.

– Make appropriate emergency DNS changes to point users to DR environment per definitions for each application environment.


Documentation

• Return Procedure:

– Confirm DISASTER OVER declaration

– Reverse disk replication; confirm synchronization

– Follow instructions for confirmation of original production environment for each application.

– Bring down DR guests identified by business units for each application environment.

– Make appropriate DNS changes to point users to non-DR environment per definitions for each application environment.

– Resume normal disk replication.


Teamwork

• Recovery coordinator• z/VM System Programmers• Linux System Administrators• Middleware

– WAS Administrators– Database Administrators– MQ Administrators

• Application Teams– Testing methodology– Expected results

Avoid processes that are dependent on subject matter experts (SME) when a disaster happens


Execution

• Test• Document results• Compare to plan• Repeat


Execution

• Where…– … to recover the systems

• Your own second site

• A recovery vendor

– … do the people go• Identify what personnel need to travel to recovery site

– Document travel procedures

• Identify alternate (local) office space

– Some office locations may be able to access recovery site if connectivity is available


Execution

• Testing– Test as often as feasible

• Frequency may depend on having your own site or contracting with a vendor

– Tests should be as close as possible to real recovery conditions

– Operating systems are easy

– Some subsystems are not so easy (e.g. large database)

– Multi-platform applications can be are more complex

– Automate as much as possible to avoid manual effort


Document ResultsCompare to Plan

• Detailed plans for all test scenarios• Carefully track tests• Document action items and follow up for improvements• Build on successes


Repeat

• Do it again• Do it regularly• Corporate emphasis may be required to encourage all

applications to test


Contact Information

Rick BarlowSenior z/VM Systems Programmer

Phone: (614) 249-5213

Internet: [email protected]

“And I thought we were busy before Linux showed up!”

mailto:[email protected]

Documents

Virtual Linux Server Disaster Recovery Planning...infrastructure to restart operations after a disaster. While many of today's larger computer systems contain built-in programs for